Abstract: The uncertainty of multi-objective temporal dependencies is difficult to be expressed or described in the modeling for recognizing the maneuverability of vehicles. To solve this problem,we propose a modeling method based on the multi-hidden Markov model (M-HMM). We model the spatio-temporal dependencies of traffic micro-situations in the influence of many factors by the method to identify and forecast the maneuvering behavior of vehicles in the traffic scenes. We apply the Baum-Welch algorithm and the forward algorithm to generate two types of input data for model training and evaluation. We also build a database of vehicles' maneuvering behaviors using a driving simulation method to improve the learning efficiency of the parameters of HMM. The experimental results of overtaking on the expressway show that at the point of time when the left tires of a vehicle passing lane the markers. The model's prediction accuracy of lane change is 98.3%,and the accuracy during the 0.4 s before true lane change is about 75%.